Why a game stutters at high FPS and how to find the cause

The counter shows plenty of FPS, yet the picture still hitches. Usually the culprit is not the average but a few long frames. Here is the difference, and a repeatable way to measure it in FPS Lab.

The short answer

Average FPS tells you how many frames come out per second on average. A stutter is a single frame that took far longer than its neighbours. A few of those barely move the average, but your eyes notice them. So look at frame times and the 1% low, and find the cause with series of captures of the same scene, changing only one thing between the series.

Average FPS and frame time

At 60 FPS each frame takes 16.7 ms on average (1000 ÷ 60). If almost every frame takes 16 ms and four take 46 ms, the average stays at the same 60 FPS, but at those four moments the picture freezes for almost three frames. Below are two runs of 200 frames with the same average.

The average is identical, while the 1% low differs almost threefold. In NyxBoost the 1% low is 1000 divided by the 99th percentile of frame time, i.e. the FPS of the slowest 1% of frames. FPS Lab also counts stutters: frames longer than twice the median and at least 8 ms over it. The even run has 0, the other one 4.

Where long frames come from

• Loading. The game compiles shaders or streams a new area from disk. These hitches often disappear the second time you play the same scene.
• Background work. Update installs, antivirus scans, Xbox Game Bar background recording, cloud sync.
• The CPU. It hits its limit or drops its clock: the power plan, a laptop's power limit, overheating.
• Frame delivery. A frame limiter, V-Sync and the monitor's refresh rate in an unlucky combination give an uneven rhythm.

No need to guess which one is yours: a measurement tells you.

A capture plan in FPS Lab

FPS Lab is a free part of NyxBoost. It gets frames from Intel PresentMon, which records when each frame is presented through Event Tracing for Windows (ETW). On the first capture NyxBoost downloads PresentMon from Intel's official page and checks the file.

1. Open Gaming → My games and pick the game on the left. The FPS LAB card is the last one on the right; the Compare performance button on the game card scrolls to it.
2. Choose a scene you can repeat: a built-in benchmark, the same route, or one spot on the map. Don't change the game's resolution or graphics: NyxBoost does not record them, so keeping them the same is up to you.
3. Select Before changes, choose the Capture length (60 s by default), start the game, get to the scene and press Measure. A run cut short with Stop is not saved.
4. Repeat at least three times: with fewer runs FPS Lab does not compare the series.
5. Make one change — for example, turn off Game Bar background recording or switch the power plan.
6. Select After changes and take three more runs of the same scene.

Reading the result

• Faster or slower. FPS Lab counts a difference only if every “after” run is faster (or slower) than every “before” run. Otherwise the verdict is “Within noise”: the difference is smaller than the spread between runs.
• “The series are not comparable”. Conditions drifted: a laptop on battery in one run and on mains in another, a new graphics driver, runs of different length, the CPU more than 30% busy before a run, or the power plan and settings changing inside one series. Take the series again.
• Same average, higher 1% low. That is fewer stutters, even if the counter in the corner did not move.
• Stutters unchanged. The change was not about your cause. Undo it — every NyxBoost setting has Revert — and test the next idea.

Under each capture FPS Lab shows an analysis: what the data suggests is limiting the frames, how confident that is, and what else could explain it.

Next step

Start with the cheapest ideas: background recording, the power plan, Game Mode. What affects FPS at all and what doesn't is in the general guide. Test each idea in its own series and keep only what FPS Lab counts as a difference.